Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Research ideation workflow from literature search to research planning. Triggers on "brainstorm research ideas", "identify research gaps", "conduct gap analysis", "start research project", "conduct literature review", "define research question", "select research method", "plan research", "research ideation".
.claude/skills/catlog22-scholar-ideation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 109% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 113% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 121% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 93% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 80% | 0% |
Supports the complete research project initiation workflow: from literature search and gap analysis through research question formulation, method selection, and research planning. Produces a structured research plan with literature review, identified gaps, formulated questions, selected methods, and a timeline.
.workflow/codebase/ARCHITECTURE.md exists, read for project contextmaestro load --type spec --category coding — load coding conventionsmaestro search "academic writing research paper" --json — top 5 entries as prior contextUser Input (research topic/interest)
|
v
┌─────────────────────────────────────────────────────────────┐
│ SKILL.md Orchestrator │
│ Collect preferences → Dispatch phases → Track progress │
└──────────┬──────────────────────────────────────────────────┘
|
┌───────┼───────┬───────────┬───────────┬───────────┐
v v v v v v
┌──────┐┌──────┐┌──────┐┌──────────┐┌──────────┐┌──────────┐
│Input ││Phase1││Phase2││ Phase3 ││ Phase4 ││ Phase5 │
│Parse ││LitSrc││GapAnl││ RQ Form ││ Method ││ Plan │
└──────┘└──────┘└──────┘└──────────┘└──────────┘└──────────┘
| | | | | |
topic papers gaps questions methods plan
scope trends opps hypotheses justify timelineBefore dispatching to phases, collect research context and workflow preferences:
AskUserQuestion:
question: "Please describe your research topic or interest area."
→ Store as: researchTopic
AskUserQuestion:
question: "What is the scope and context of your research?"
options:
- "Broad exploration (survey a field)"
- "Focused investigation (specific problem)"
- "Applied research (practical application)"
→ Store as: researchScope
AskUserQuestion:
question: "What is your target research timeline?"
options:
- "Short-term (3-6 months)"
- "Medium-term (6-12 months)"
- "Long-term (1-2 years)"
→ Store as: researchTimeline
AskUserQuestion:
question: "Do you have access to Zotero for literature management?"
options:
- "Yes (auto-import papers to Zotero)"
- "No (skip Zotero integration)"
→ Store as: useZotero
AskUserQuestion:
question: "Workflow mode?"
options:
- "Interactive (confirm at each phase)"
- "Auto (run all phases continuously)"
→ Store as: workflowModeDerived preferences:
workflowPreferences = {
topic: researchTopic,
scope: researchScope,
timeline: researchTimeline,
useZotero: useZotero === "Yes",
autoYes: workflowMode === "Auto"
}When workflowPreferences.autoYes === true: Execute all 5 phases sequentially without confirmation prompts between phases. Still pause for user input when phase-specific questions arise (e.g., confirming search keywords, selecting papers for deep reading).
> COMPACT DIRECTIVE: Context compression MUST check TodoWrite phase status. > The phase currently marked in_progress is the active execution phase — preserve its FULL content. > Only compress phases marked completed or pending.
TodoWrite:
- "Phase 1: Literature Search" (pending)
- "Phase 2: Gap Analysis" (pending)
- "Phase 3: Research Question Formulation" (pending)
- "Phase 4: Method Selection" (pending)
- "Phase 5: Research Planning" (pending)Phase 1: Literature Search
Mark TodoWrite Phase 1 → in_progress
└─ Ref: phases/01-literature-search.md
├─ Input: workflowPreferences (topic, scope, useZotero)
└─ Output: literatureResults (papers, trends, keyFindings)
Phase 2: Gap Analysis
Mark TodoWrite Phase 1 → completed, Phase 2 → in_progress
└─ Ref: phases/02-gap-analysis.md
├─ Input: literatureResults
└─ Output: gapAnalysis (gaps, opportunities, priorities)
Phase 3: Research Question Formulation
Mark TodoWrite Phase 2 → completed, Phase 3 → in_progress
└─ Ref: phases/03-research-question.md
├─ Input: gapAnalysis + literatureResults
└─ Output: researchQuestions (questions, hypotheses, objectives)
Phase 4: Method Selection
Mark TodoWrite Phase 3 → completed, Phase 4 → in_progress
└─ Ref: phases/04-method-selection.md
├─ Input: researchQuestions + gapAnalysis
└─ Output: selectedMethods (methods, justification, resources)
Phase 5: Research Planning
Mark TodoWrite Phase 4 → completed, Phase 5 → in_progress
└─ Ref: phases/05-research-planning.md
├─ Input: ALL previous outputs
└─ Output: research-plan.md (final deliverable)
Mark TodoWrite Phase 5 → completedPhase Reference Documents (read on-demand when phase executes):
| Phase | Document | Purpose | Compact | |-------|----------|---------|---------| | 1 | phases/01-literature-search.md | Search, discover, and organize literature | TodoWrite driven | | 2 | phases/02-gap-analysis.md | Identify research gaps and opportunities | TodoWrite driven | | 3 | phases/03-research-question.md | Formulate research questions and hypotheses | TodoWrite driven | | 4 | phases/04-method-selection.md | Select and justify research methods | TodoWrite driven + sentinel | | 5 | phases/05-research-planning.md | Create timeline, milestones, and final plan | TodoWrite driven + sentinel |
Compact Rules:
in_progress — preserve full content, do not compresscompleted — may compress to summaryRead() to recoverUser input (free text or structured) is converted to:
TOPIC: [research topic or interest]
SCOPE: [broad | focused | applied]
TIMELINE: [short | medium | long]
CONTEXT: [additional constraints, domain, background]If user provides a simple topic string, derive scope and timeline from the topic description or ask via AskUserQuestion.
workflowPreferences
├─→ Phase 1: topic, scope, useZotero
│ └─→ literatureResults {papers[], trends[], keyFindings[], zoteroCollection?}
│
├─→ Phase 2: literatureResults
│ └─→ gapAnalysis {gaps[], opportunities[], prioritizedGaps[]}
│
├─→ Phase 3: gapAnalysis + literatureResults
│ └─→ researchQuestions {questions[], hypotheses[], objectives[]}
│
├─→ Phase 4: researchQuestions + gapAnalysis
│ └─→ selectedMethods {methods[], justification, resources[]}
│
└─→ Phase 5: ALL outputs
└─→ research-plan.md (final structured document)Phase start — Attach sub-tasks:
Mark Phase N → in_progress
Add sub-tasks:
- "Step N.1: ..." (in_progress)
- "Step N.2: ..." (pending)
- "Step N.3: ..." (pending)Phase end — Collapse:
Mark all Phase N sub-tasks → completed
Mark Phase N → completed
Mark Phase N+1 → in_progressAfter each phase completes, update an accumulated research notes document:
markdown## Research Notes (accumulated) ### After Phase 1 (Literature Search) - Papers found: [count] - Key trends: [list] - Zotero collection: [name, if applicable] ### After Phase 2 (Gap Analysis) - Gaps identified: [count] - Top priorities: [list] ### After Phase 3 (Research Questions) - Primary RQ: [question] - Sub-questions: [count] ### After Phase 4 (Method Selection) - Selected approach: [method] - Required resources: [list]
Pre-phase:
Read("phases/0N-xxx.md"))Post-phase:
The workflow produces:
research-plan.md — Final structured research proposal with all sectionsliterature-review.md — Structured literature review with categorized papers| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 30,417 | 6,502 | -79% | 1 | 1 | 0% | 4,500 | 3,715 | -17% | 0 | 0 | — |
case-02 | fail→fail | 39,106 | 3,928 | -90% | 1 | 1 | 0% | 6,207 | 3,335 | -46% | 0 | 0 | — |
case-18 | pass→pass | 12,566 | 3,882 | -69% | 1 | 1 | 0% | 1,964 | 3,325 | +69% | 0 | 0 | — |
case-03 | fail→fail | 39,487 | 11,950 | -70% | 1 | 1 | 0% | 6,200 | 3,637 | -41% | 0 | 0 | — |
case-04 | fail→fail | 26,898 | 30,891 | +15% | 1 | 1 | 0% | 4,975 | 8,904 | +79% | 0 | 0 | — |
case-05 | fail→fail | 29,257 | 20,452 | -30% | 1 | 1 | 0% | 4,444 | 6,304 | +42% | 0 | 0 | — |
case-06 | fail→fail | 26,212 | 35,174 | +34% | 1 | 1 | 0% | 4,433 | 8,702 | +96% | 0 | 0 | — |
case-07 | pass→fail | 13,969 | 12,473 | -11% | 1 | 1 | 0% | 2,058 | 4,514 | +119% | 0 | 0 | — |
case-08 | pass→pass | 7,014 | 3,626 | -48% | 1 | 1 | 0% | 1,093 | 3,214 | +194% | 0 | 0 | — |
case-09 | fail→pass | 15,637 | 11,875 | -24% | 1 | 1 | 0% | 2,082 | 4,357 | +109% | 0 | 0 | — |
case-10 | fail→pass | 12,936 | 9,952 | -23% | 1 | 1 | 0% | 1,916 | 4,073 | +113% | 0 | 0 | — |
case-11 | pass→pass | 15,091 | 16,205 | +7% | 1 | 1 | 0% | 2,067 | 4,805 | +132% | 0 | 0 | — |
case-12 | fail→pass | 12,629 | 7,638 | -40% | 1 | 1 | 0% | 1,722 | 3,810 | +121% | 0 | 0 | — |
case-13 | fail→pass | 11,188 | 2,334 | -79% | 1 | 1 | 0% | 1,556 | 3,002 | +93% | 0 | 0 | — |
case-14 | pass→pass | 13,659 | 3,328 | -76% | 1 | 1 | 0% | 1,855 | 3,192 | +72% | 0 | 0 | — |
case-15 | fail→pass | 16,487 | 8,981 | -46% | 1 | 1 | 0% | 2,188 | 3,946 | +80% | 0 | 0 | — |
case-16 | fail→pass | 15,868 | 6,189 | -61% | 1 | 1 | 0% | 2,190 | 3,544 | +62% | 0 | 0 | — |
case-17 | pass→pass | 18,823 | 15,384 | -18% | 1 | 1 | 0% | 2,731 | 4,757 | +74% | 0 | 0 | — |
case-19 | fail→fail | 11,394 | 2,314 | -80% | 1 | 1 | 0% | 1,670 | 3,025 | +81% | 0 | 0 | — |
case-20 | fail→fail | 14,132 | 4,008 | -72% | 1 | 1 | 0% | 1,949 | 3,339 | +71% | 0 | 0 | — |
case-21 | fail→pass | 12,277 | 4,292 | -65% | 1 | 1 | 0% | 1,665 | 3,389 | +104% | 0 | 0 | — |
case-22 | fail→pass | 9,258 | 4,069 | -56% | 1 | 1 | 0% | 1,429 | 3,282 | +130% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +32 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.